Hollywood’s AI problem is no longer adoption. It is who controls the hidden workflow
A new WIRED interview with entertainment-industry journalist Matthew Belloni describes AI as increasingly routine across Hollywood, including creative development, animation and production.
Belloni says public attitudes often differ from private practice and that studios, talent representatives and technology companies remain divided over copyright, employment and control of name, image and likeness. He also discusses efforts such as the Human Consent Registry and industry pressure to replace default opt-out systems with permission-based use of performers and protected material.
Why visual creators should care
A project may use AI substantially without that use appearing in the credits, production notes or finished frame. That creates uncertainty for performers, cinematographers, visual-effects artists and clients who need to know whether work, likenesses or performances entered a generative system.
Practical takeaway
Add AI disclosure to production paperwork before work begins. Record which departments may use generative tools, whether unpublished material may be uploaded, how performers’ likenesses may be processed and whether the production must disclose AI use to talent, unions, insurers or distributors.
What to watch next
Watch whether talent agencies, studios and unions converge around opt-in likeness registries or continue relying on project-by-project agreements and state publicity laws.
Sources
Reuters is putting AI interpretation inside the live-video editing window
Reuters has launched AI-generated, time-coded English transcripts for live video on Reuters Connect. Agency customers can read rolling transcripts alongside an event, identify moments and create clips in real time. When the event ends, the complete transcript remains connected to the on-demand video asset. The feature joins existing automated transcripts, translation and scene lists inside Reuters Connect.
Why visual creators should care
AI is no longer entering the editorial process only after footage has been recorded. It now influences how live material is searched, described, clipped and published while the event is still happening. An incorrect name, quotation or interpretation can therefore move into headlines and captions very quickly.
Practical takeaway
Treat the live transcript as a navigation aid rather than a verified quotation. Check important wording against the audio, record any corrections and retain the time-coded source video supporting the published clip.
What to watch next
Watch whether Reuters adds confidence indicators, additional languages and clearer records showing which captions, scripts or clips were created from machine transcription rather than human logging.
Sources
Lightroom can now invent the part of the photograph that was never inside the frame
Adobe Lightroom version 9.5 adds Generative Expand, allowing a photographer to enlarge the frame and generate new content beyond the original captured boundaries. The tool is presented as part of Lightroom’s normal desktop editing workflow rather than a separate image-generation application.
Why visual creators should care
A conventional crop removes captured information. Generative expansion does the opposite: it adds plausible visual material the camera never recorded. That distinction can be important for documentary work, property imagery, product photography, competitions and any client expecting the delivered photograph to represent an actual scene.
Practical takeaway
Keep the uncropped camera original and record when Generative Expand was used. For factual, architectural or product work, disclose expansion whenever the generated area could change how the subject, space or product is understood.
What to watch next
Watch whether Lightroom exposes generated-region history and Content Credentials clearly after export and whether stock agencies and photography competitions publish specific rules for expanded photographs.
Sources
A rendered DNG can preserve the look while hiding how the photograph reached it
Lightroom Classic 15.5 introduces Render to DNG, which creates a new DNG file with the photograph’s current edits permanently applied. Adobe describes it as a non-raw file that provides a clean starting point for additional work. The same release adds masking controls, improved crop visibility and other production features.
Why visual creators should care
A rendered derivative may look like a DNG while no longer functioning as the untouched camera raw file. Once edits are baked in, later recipients may be unable to distinguish camera data from retouching, masking, denoising or AI-assisted changes by looking only at the new file.
Practical takeaway
Never replace the original raw file with the rendered DNG. Store both, label the derivative clearly and preserve the Lightroom catalog or sidecar information showing which edits were applied before rendering.
What to watch next
Watch how Render to DNG records generative and AI edits in metadata and how authenticity systems distinguish rendered derivatives from camera-originated DNG files.
Sources
Generative expansion is moving closer to the raw-processing stage
Adobe Camera Raw version 18.4.1 adds Generative Expand, allowing photographers to extend the canvas and fill new areas with generated content before the image enters many later Photoshop or publishing workflows. The release also adds a fullscreen presentation mode.
Why visual creators should care
When a synthetic alteration happens close to raw processing, later collaborators may assume the expanded image is simply a developed photograph. The earlier AI enters the chain, the easier it is for the transformation to disappear from the project’s visible history.
Practical takeaway
Save an untouched raw original and a separate AI-expanded version. Name the derivative clearly and make the expansion part of the handoff notes when the file moves to a retoucher, designer, editor or client.
What to watch next
Watch whether Adobe standardises the way Lightroom, Camera Raw and Photoshop expose the same generative operation in Content Credentials and editing metadata.
Sources
Indonesia’s draft would protect human AI-assisted work while prohibiting distinctive-style imitation
Indonesia is considering a copyright overhaul that would explicitly address AI-assisted creation, training and disclosure. According to a draft reviewed by Reuters, sufficiently human AI-assisted works could qualify for copyright while fully generated works would not. The proposal would prohibit AI imitation of a creator’s distinctive style, require disclosure of AI use and make some training uses subject to fair-use rules or licensing. The provisions would apply to photography, film, journalism, video games and other protected media.
Why visual creators should care
The proposal combines several issues that other countries often treat separately: human authorship, training compensation, disclosure and imitation of an identifiable creative style. For visual creators, it could provide a route to challenge commercial mimicry even when no single finished image has been copied exactly.
Practical takeaway
Creators working in or licensing into Indonesia should preserve evidence of human contribution, distinctive bodies of work and authorised training or reference uses. The proposal remains a draft, so it should not yet be treated as current law.
What to watch next
Watch the final human-contribution threshold, the definition of distinctive style, available remedies and whether the training-compensation provisions survive industry consultation.
Sources
